Applying polynomial decoupling methods to the polynomial NARX model

System identification uses measurements of a dynamic system's input and output to reconstruct mathematical model for that system. These can be mechanical, electrical, physiological, among others. Since most the systems around us exhibit some form nonlinear behavior, system techniques are tools will help gain better understanding our surroundings potentially let improve their performance. One is often used represent polynomial NARX model, an equation error where function past inputs outputs. That said, major disadvantage with number parameters increases rapidly increasing order. Furthermore, black-box therefore difficult interpret. This paper discusses decoupling algorithm substitutes multivariate transformation matrix followed by bank univariate polynomials. decreases significantly also imposes structure on model. non-convex optimization required this technique, initialization important factor consider. In developed in conjunction several different techniques. The resulting algorithms applied two benchmark problems: measurement data from Silver-Box simulation Bouc-Wen friction performance evaluated validation signals both prediction.